@InProceedings{Filippozzi2024,
  author    = {Lucia Filippozzi and Santiago Mazuelas and I{\~{n}}igo Urteaga},
  booktitle = {Proceedings of the 10th Machine Learning for Healthcare},
  title     = {{Minimax Risk Classifiers for Mislabeled Data: a Study on Patient Outcome Prediction Tasks}},
  year      = {2024},
  month     = {16--17 Aug},
  pages     = {1--52},
  publisher = {PMLR},
  series    = {Proceedings of Machine Learning Research},
  volume    = {252},
  url       = {https://proceedings.mlr.press/v252/filippozzi24a.html},
}

@InProceedings{Urteaga2023,
  author    = {I{\~n}igo Urteaga and Moulay-Za\"idane Dra\"idia and Tomer Lancewicki and Shahram Khadivi},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2023},
  title     = {{Multi-armed bandits for resource efficient, online optimization of language model pre-training: the use case of dynamic masking}},
  year      = {2023},
  address   = {Toronto, Canada},
  month     = jul,
  pages     = {10609--10627},
  publisher = {Association for Computational Linguistics},
  abstract  = {We design and evaluate a Bayesian optimization framework for resource efficient pre-training of Transformer-based language models (TLMs). TLM pre-training requires high computational resources and introduces many unresolved design choices, such as selecting its pre-training hyperparameters.We propose a multi-armed bandit framework for the sequential selection of pre-training hyperparameters, aimed at optimizing language model performance, in a resource efficient manner. We design a Thompson sampling algorithm, with a surrogate Gaussian process reward model of the Masked Language Model (MLM) pre-training objective, for its sequential minimization. Instead of MLM pre-training with fixed masking probabilities, the proposed Gaussian process-based Thompson sampling (GP-TS) accelerates pre-training by sequentially selecting masking hyperparameters that improve performance.We empirically demonstrate how GP-TS pre-trains language models efficiently, i.e., it achieves lower MLM loss in fewer epochs, across a variety of settings. In addition, GP-TS pre-trained TLMs attain competitive downstream performance, while avoiding expensive hyperparameter grid search. GP-TS provides an interactive framework for efficient and optimized TLM pre-training that, by circumventing costly hyperparameter selection, enables substantial computational savings.},
  doi       = {10.18653/v1/2023.findings-acl.675},
  url       = {https://aclanthology.org/2023.findings-acl.675},
}

We design and evaluate a Bayesian optimization framework for resource efficient pre-training of Transformer-based language models (TLMs). TLM pre-training requires high computational resources and introduces many unresolved design choices, such as selecting its pre-training hyperparameters.We propose a multi-armed bandit framework for the sequential selection of pre-training hyperparameters, aimed at optimizing language model performance, in a resource efficient manner. We design a Thompson sampling algorithm, with a surrogate Gaussian process reward model of the Masked Language Model (MLM) pre-training objective, for its sequential minimization. Instead of MLM pre-training with fixed masking probabilities, the proposed Gaussian process-based Thompson sampling (GP-TS) accelerates pre-training by sequentially selecting masking hyperparameters that improve performance.We empirically demonstrate how GP-TS pre-trains language models efficiently, i.e., it achieves lower MLM loss in fewer epochs, across a variety of settings. In addition, GP-TS pre-trained TLMs attain competitive downstream performance, while avoiding expensive hyperparameter grid search. GP-TS provides an interactive framework for efficient and optimized TLM pre-training that, by circumventing costly hyperparameter selection, enables substantial computational savings.

@InProceedings{ip-Urteaga2021,
  author    = {I{\~{n}}igo Urteaga and Kathy Li and Chris Wiggins and No{\'{e}}mie Elhadad},
  booktitle = {Proceedings of the 6th Machine Learning for Healthcare Conference},
  title     = {{A Generative Modeling Approach to Calibrated Predictions: A Use Case on Menstrual Cycle Length Prediction}},
  year      = {2021},
  editor    = {Jung, Ken and Yeung, Serena and Sendak, Mark and Sjoding, Michael and Ranganath, Rajesh},
  month     = {06--07 Aug},
  pages     = {535--566},
  publisher = {PMLR},
  series    = {Proceedings of Machine Learning Research},
  volume    = {149},
  abstract  = {We present an end-to-end statistical framework for personalized, accurate, and minimally invasive modeling of female reproductive hormonal patterns. Reconstructing and forecasting the evolution of hormonal dynamics is a challenging task, but a critical one to improve general understanding of the menstrual cycle and personalized detection of potential health issues. Our goal is to infer and forecast individual hormone daily levels over time, while accommodating pragmatic and minimally invasive measurement settings. To that end, our approach combines the power of probabilistic generative models (i.e., multi-task Gaussian processes) with the flexibility of neural networks (i.e., a dilated convolutional architecture) to learn complex temporal mappings. To attain accurate hormone level reconstruction with as little data as possible, we propose a sampling mechanism for optimal reconstruction accuracy with limited sampling budget. Our results show the validity of our proposed hormonal dynamic modeling framework, as it provides accurate predictive performance across different realistic sampling budgets and outperforms baselines methods.},
  file      = {urteaga19a.pdf:http\://proceedings.mlr.press/v106/urteaga19a/urteaga19a.pdf:PDF},
  url       = {https://proceedings.mlr.press/v149/urteaga21a},
}

We present an end-to-end statistical framework for personalized, accurate, and minimally invasive modeling of female reproductive hormonal patterns. Reconstructing and forecasting the evolution of hormonal dynamics is a challenging task, but a critical one to improve general understanding of the menstrual cycle and personalized detection of potential health issues. Our goal is to infer and forecast individual hormone daily levels over time, while accommodating pragmatic and minimally invasive measurement settings. To that end, our approach combines the power of probabilistic generative models (i.e., multi-task Gaussian processes) with the flexibility of neural networks (i.e., a dilated convolutional architecture) to learn complex temporal mappings. To attain accurate hormone level reconstruction with as little data as possible, we propose a sampling mechanism for optimal reconstruction accuracy with limited sampling budget. Our results show the validity of our proposed hormonal dynamic modeling framework, as it provides accurate predictive performance across different realistic sampling budgets and outperforms baselines methods.

@InProceedings{ip-Urteaga2019,
  author    = {I{\~{n}}igo Urteaga and Tristan Bertin and Theresa M. Hardy and David J. Albers and No{\'{e}}mie Elhadad},
  title     = {{Multi-Task Gaussian Processes and Dilated Convolutional Networks for Reconstruction of Reproductive Hormonal Dynamics}},
  booktitle = {Proceedings of the 4th Machine Learning for Healthcare},
  year      = {2019},
  volume    = {106},
  series    = {Proceedings of Machine Learning Research},
  pages     = {66--90},
  month     = {09--10 Aug},
  publisher = {PMLR},
  abstract  = {We present an end-to-end statistical framework for personalized, accurate, and minimally invasive modeling of female reproductive hormonal patterns. Reconstructing and forecasting the evolution of hormonal dynamics is a challenging task, but a critical one to improve general understanding of the menstrual cycle and personalized detection of potential health issues. Our goal is to infer and forecast individual hormone daily levels over time, while accommodating pragmatic and minimally invasive measurement settings. To that end, our approach combines the power of probabilistic generative models (i.e., multi-task Gaussian processes) with the flexibility of neural networks (i.e., a dilated convolutional architecture) to learn complex temporal mappings. To attain accurate hormone level reconstruction with as little data as possible, we propose a sampling mechanism for optimal reconstruction accuracy with limited sampling budget. Our results show the validity of our proposed hormonal dynamic modeling framework, as it provides accurate predictive performance across different realistic sampling budgets and outperforms baselines methods.},
  file      = {urteaga19a.pdf:http\://proceedings.mlr.press/v106/urteaga19a/urteaga19a.pdf:PDF},
  url       = {http://proceedings.mlr.press/v106/urteaga19a.html},
}

We present an end-to-end statistical framework for personalized, accurate, and minimally invasive modeling of female reproductive hormonal patterns. Reconstructing and forecasting the evolution of hormonal dynamics is a challenging task, but a critical one to improve general understanding of the menstrual cycle and personalized detection of potential health issues. Our goal is to infer and forecast individual hormone daily levels over time, while accommodating pragmatic and minimally invasive measurement settings. To that end, our approach combines the power of probabilistic generative models (i.e., multi-task Gaussian processes) with the flexibility of neural networks (i.e., a dilated convolutional architecture) to learn complex temporal mappings. To attain accurate hormone level reconstruction with as little data as possible, we propose a sampling mechanism for optimal reconstruction accuracy with limited sampling budget. Our results show the validity of our proposed hormonal dynamic modeling framework, as it provides accurate predictive performance across different realistic sampling budgets and outperforms baselines methods.

@InProceedings{ip-Urteaga2018,
  author    = {I{\~n}igo Urteaga and Mollie McKillop and Sharon Lipsky-Gorman and No{\'e}mie Elhadad},
  title     = {{Phenotyping Endometriosis through Mixed Membership Models of Self-Tracking Data}},
  booktitle = {2018 Machine Learning for Healthcare (MLHC)},
  year      = {2018},
  owner     = {iurteaga},
  timestamp = {2017-04-07},
  url       = {https://www.mlforhc.org/s/27.pdf},
}

@InProceedings{j-Urteaga2018a,
  author    = {I{\~n}igo Urteaga and Chris Wiggins},
  title     = {{Variational inference for the multi-armed contextual bandit}},
  booktitle = {Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics},
  year      = {2018},
  volume    = {84},
  series    = {Proceedings of Machine Learning Research},
  pages     = {698--706},
  month     = {09--11 Apr},
  publisher = {PMLR},
  abstract  = {In many biomedical, science, and engineering problems, one must sequentially decide which action to take next so as to maximize rewards. One general class of algorithms for optimizing interactions with the world, while simultaneously learning how the world operates, is the multi-armed bandit setting and, in particular, the contextual bandit case. In this setting, for each executed action, one observes rewards that are dependent on a given ’context’, available at each interaction with the world. The Thompson sampling algorithm has recently been shown to enjoy provable optimality properties for this set of problems, and to perform well in real-world settings. It facilitates  generative and interpretable modeling of the problem at hand. Nevertheless, the design and complexity of the model limit its application, since one must both sample from the distributions modeled and calculate their expected rewards. We here show how these limitations can be overcome using variational inference to approximate complex models, applying to the reinforcement learning case advances developed for the inference case in the machine learning community over the past two decades. We consider contextual multi-armed bandit applications where the true reward distribution is unknown and complex, which we approximate with a mixture model whose parameters are inferred via variational inference. We show how the proposed variational Thompson sampling approach is accurate in approximating the true distribution, and attains reduced regrets even with complex reward distributions. The proposed algorithm is valuable for practical scenarios where restrictive modeling assumptions are undesirable.},
  file      = {urteaga18a.pdf:http\://proceedings.mlr.press/v84/urteaga18a/urteaga18a.pdf:PDF},
  url       = {http://proceedings.mlr.press/v84/urteaga18a.html},
}

In many biomedical, science, and engineering problems, one must sequentially decide which action to take next so as to maximize rewards. One general class of algorithms for optimizing interactions with the world, while simultaneously learning how the world operates, is the multi-armed bandit setting and, in particular, the contextual bandit case. In this setting, for each executed action, one observes rewards that are dependent on a given ’context’, available at each interaction with the world. The Thompson sampling algorithm has recently been shown to enjoy provable optimality properties for this set of problems, and to perform well in real-world settings. It facilitates generative and interpretable modeling of the problem at hand. Nevertheless, the design and complexity of the model limit its application, since one must both sample from the distributions modeled and calculate their expected rewards. We here show how these limitations can be overcome using variational inference to approximate complex models, applying to the reinforcement learning case advances developed for the inference case in the machine learning community over the past two decades. We consider contextual multi-armed bandit applications where the true reward distribution is unknown and complex, which we approximate with a mixture model whose parameters are inferred via variational inference. We show how the proposed variational Thompson sampling approach is accurate in approximating the true distribution, and attains reduced regrets even with complex reward distributions. The proposed algorithm is valuable for practical scenarios where restrictive modeling assumptions are undesirable.

@InProceedings{ip-Urteaga2017,
  author    = {I{\~n}igo Urteaga and Petar M Djuri\'{c}},
  title     = {{Multiple Particle Filtering for Inference in the presence of state correlation of unknown mixing parameters}},
  booktitle = {2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  year      = {2017},
  pages     = {3849--3853},
  owner     = {iurteaga},
  timestamp = {2017-04-07},
}

@InProceedings{ip-Urteaga2016,
  author    = {I{\~n}igo Urteaga and M\'{o}nica F. Bugallo and Petar M Djuri\'{c}},
  title     = {{Sequential Monte Carlo methods under model uncertainty}},
  booktitle = {2016 IEEE Statistical Signal Processing Workshop (SSP)},
  year      = {2016},
  pages     = {1-5},
  month     = {June},
  abstract  = {We propose a Sequential Monte Carlo (SMC) method for filtering and prediction of time-varying signals under model uncertainty. Instead of resorting to model selection, we fuse the information from the considered models within the proposed SMC method. We achieve our goal by dynamically adjusting the resampling step according to the posterior predictive power of each model, which is updated sequentially as we observe more data. The method allows the models with better predictive powers to explore the state space with more resources than models lacking predictive power. This is done autonomously and dynamically within the SMC method. We show the validity of the presented method by evaluating it on an illustrative application.},
  doi       = {10.1109/SSP.2016.7551747},
  keywords  = {Monte Carlo methods;filtering theory;prediction theory;signal sampling;state-space methods;SMC method;information fusion;model uncertainty;resampling step;sequential Monte Carlo method;state-space method;time-varying signal filtering;time-varying signal prediction;Atmospheric measurements;Computational modeling;Mathematical model;Monte Carlo methods;Particle measurements;Predictive models;Uncertainty;Sequential Monte Carlo;dynamic model averaging;information fusion;particle filtering;resampling},
  owner     = {iurteaga},
  timestamp = {2016-04-29},
}

We propose a Sequential Monte Carlo (SMC) method for filtering and prediction of time-varying signals under model uncertainty. Instead of resorting to model selection, we fuse the information from the considered models within the proposed SMC method. We achieve our goal by dynamically adjusting the resampling step according to the posterior predictive power of each model, which is updated sequentially as we observe more data. The method allows the models with better predictive powers to explore the state space with more resources than models lacking predictive power. This is done autonomously and dynamically within the SMC method. We show the validity of the presented method by evaluating it on an illustrative application.

@InProceedings{ip-Urteaga2016a,
  author    = {I{\~n}igo Urteaga and M\'{o}nica F. Bugallo and Petar M Djuri\'{c}},
  title     = {{Sequential Monte Carlo sampling for correlated latent long-memory time-series}},
  booktitle = {2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  year      = {2016},
  pages     = {6580-6584},
  month     = {March},
  abstract  = {In this paper, we consider state-space models where the latent processes represent correlated mixtures of fractional Gaussian processes embedded in white Gaussian noises. The observed data are nonlinear functions of the latent states. The fractional Gaussian processes have interesting properties including long-memory, self-similarity and scale-invariance, and thus, are of interest for building models in finance and econometrics. We propose sequential Monte Carlo (SMC) methods for inference of the latent processes where each method is based on different assumptions about the parameters of the state-space model. The methods are extensively evaluated via simulations of the popular stochastic volatility model.},
  doi       = {10.1109/ICASSP.2016.7472945},
  keywords  = {Gaussian noise;Monte Carlo methods;particle filtering (numerical methods);sequential estimation;state-space methods;stochastic processes;time series;correlated latent long-memory time-series;correlated mixtures;fractional Gaussian processes;nonlinear functions;sequential Monte Carlo sampling;state-space models;stochastic volatility model;white Gaussian noises;Biological system modeling;Econometrics;Gaussian noise;Gaussian processes;Monte Carlo methods;State-space methods;Sequential Monte Carlo;operator fractional Gaussian process;particle filtering;state-space models;time-series},
  owner     = {iurteaga},
  timestamp = {2016-04-14},
}

In this paper, we consider state-space models where the latent processes represent correlated mixtures of fractional Gaussian processes embedded in white Gaussian noises. The observed data are nonlinear functions of the latent states. The fractional Gaussian processes have interesting properties including long-memory, self-similarity and scale-invariance, and thus, are of interest for building models in finance and econometrics. We propose sequential Monte Carlo (SMC) methods for inference of the latent processes where each method is based on different assumptions about the parameters of the state-space model. The methods are extensively evaluated via simulations of the popular stochastic volatility model.

@InCollection{ic-Perez-Sanchez2015,
  Title                    = {{DTN Routing Optimised by Human Routines: The HURRy Protocol}},
  Author                   = {Susana P{\'e}rez-S{\'a}nchez and Jos{\'e} Mar{\'i}a Cabero and I{\~n}igo Urteaga},
  Booktitle                = {Wired/Wireless Internet Communications},
  Publisher                = {Springer International Publishing},
  Year                     = {2015},
  Pages                    = {299-312},
  Series                   = {Lecture Notes in Computer Science},
  Volume                   = {9071},

  Abstract                 = {This paper proposes the HURRy (HUman Routines used for Routing) protocol, which infers and benefits from the social behaviour of nodes in disruptive networking environments. HURRy incorporates the contact duration to the information retrieved from historical encounters among neighbours, so that smarter routing decisions can be made. The specification of HURRy is based on the outcomes of a thorough experiment, which highlighted the importance of distinguishing between short and long contacts and deriving mathematical relations in order to optimally prioritize the available routes to a destination. HURRy introduces a novel and more meaningful rating system to evaluate the quality of each contact and overcome the limitations of other routing approaches in social environments.},
  Doi                      = {10.1007/978-3-319-22572-2_22},
  ISBN                     = {978-3-319-22571-5},
  Keywords                 = {Challenged networks; DTN; Probabilistic routing; Social behaviour},
  Language                 = {English},
  Owner                    = {iurteaga},
  Timestamp                = {2015-11-10}
}

This paper proposes the HURRy (HUman Routines used for Routing) protocol, which infers and benefits from the social behaviour of nodes in disruptive networking environments. HURRy incorporates the contact duration to the information retrieved from historical encounters among neighbours, so that smarter routing decisions can be made. The specification of HURRy is based on the outcomes of a thorough experiment, which highlighted the importance of distinguishing between short and long contacts and deriving mathematical relations in order to optimally prioritize the available routes to a destination. HURRy introduces a novel and more meaningful rating system to evaluate the quality of each contact and overcome the limitations of other routing approaches in social environments.

@InProceedings{ip-Urteaga2015a,
  author    = {I{\~n}igo Urteaga and M\'{o}nica F. Bugallo and Petar M Djuri\'{c}},
  title     = {{Sequential Monte Carlo sampling for systems with fractional Gaussian processes}},
  booktitle = {2015 Proceedings of the 23th European Signal Processing Conference (EUSIPCO)},
  year      = {2015},
  pages     = {1246--1250},
  doi       = {http://dx.doi.org/10.1109/EUSIPCO.2015.7362583},
  owner     = {iurteaga},
  timestamp = {2015-06-26},
}

@InProceedings{ip-Urteaga2015b,
  author    = {I{\~n}igo Urteaga and M\'{o}nica F. Bugallo and Petar M Djuri\'{c}},
  title     = {{Filtering of nonlinear time-series coupled by fractional Gaussian processes}},
  booktitle = {2015 IEEE 6th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP)},
  year      = {2015},
  pages     = {489--492},
  abstract  = {In this paper we consider a set of time\-/series that are coupled by latent fractional Gaussian processes. Specifically, we address time\-/series that combine idiosyncratic short\-/term and shared long\-/term features. The long\-/memory is modeled by fractional Gaussian processes, whereas the short\-/memory properties are captured by linear models of past data. The observations are nonlinear functions of the hidden states and therefore we resort to a sequential Monte Carlo sampling technique for inference of the latent states. The proposed solution is evaluated via simulations of an illustrative practical scenario.},
  doi       = {http://dx.doi.org/10.1109/CAMSAP.2015.7383843},
  owner     = {iurteaga},
  timestamp = {2015-11-10},
}

In this paper we consider a set of time­/series that are coupled by latent fractional Gaussian processes. Specifically, we address time­/series that combine idiosyncratic short­/term and shared long­/term features. The long­/memory is modeled by fractional Gaussian processes, whereas the short­/memory properties are captured by linear models of past data. The observations are nonlinear functions of the hidden states and therefore we resort to a sequential Monte Carlo sampling technique for inference of the latent states. The proposed solution is evaluated via simulations of an illustrative practical scenario.

@InProceedings{ip-Urteaga2015,
  Title                    = {{Particle filtering of {ARMA} processes of unknown order and parameters}},
  Author                   = {I{\~n}igo Urteaga and Petar M Djuri\'{c}},
  Booktitle                = {2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  Year                     = {2015},
  Month                    = {April},
  Pages                    = {4105-4109},

  Abstract                 = {This paper considers inference on the widely used state-space models described by hidden ARMA state processes of unknown order observed via non-linear functions of the states. We propose a particle filtering method for sequentially inferring the unknown ARMA time-series by Rao-Blackwellization of all the static unknowns. Our method does not rely either on any assumption on the model order or on the static ARMA and state innovation parameters. Consequently, when the ARMA model order is unknown, it can be used without a follow-up model selection procedure. Extensive simulation results validate the proposed method across different ARMA models.},
  Doi                      = {10.1109/ICASSP.2015.7178743},
  Keywords                 = {particle filtering (numerical methods);time series;ARMA time-series;Rao-Blackwellization;hidden ARMA state processes;innovation parameters;nonlinear functions;particle filtering method;state-space models;Computational modeling;Covariance matrices;Estimation;Mathematical model;Noise;State-space methods;Technological innovation;ARMA models;Rao-Blackwellization;State-space models;particle filtering;time-series},
  Owner                    = {iurteaga},
  Timestamp                = {2015-10-15}
}

This paper considers inference on the widely used state-space models described by hidden ARMA state processes of unknown order observed via non-linear functions of the states. We propose a particle filtering method for sequentially inferring the unknown ARMA time-series by Rao-Blackwellization of all the static unknowns. Our method does not rely either on any assumption on the model order or on the static ARMA and state innovation parameters. Consequently, when the ARMA model order is unknown, it can be used without a follow-up model selection procedure. Extensive simulation results validate the proposed method across different ARMA models.

@InProceedings{ip-Urteaga2014,
  Title                    = {{Estimation of ARMA state processes by particle filtering}},
  Author                   = {I{\~n}igo Urteaga and Petar M Djuri\'{c}},
  Booktitle                = {2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  Year                     = {2014},
  Month                    = {May},
  Pages                    = {8033-8037},

  Abstract                 = {There are many practical signal processing settings where a state-space model consists of a state described by an ARMA process that is observed via non-linear functions of the state. In this paper, we propose a particle filtering method for sequentially estimating the ARMA process in the presence of unknown parameters. In the considered problem, we have static and dynamic unknowns, and we show how to handle the static parameters so that the estimation of the state process does not degrade with time. We propose a new particle filter that approximates the posterior of all the unknowns by a Gaussian distribution, in combination with a Monte Carlo approach to the Rao-Blackwellization of the static parameters. We demonstrate the performance of the proposed method by extensive computer simulations.},
  Doi                      = {10.1109/ICASSP.2014.6855165},
  Keywords                 = {Gaussian distribution;Monte Carlo methods;autoregressive moving average processes;parameter estimation;particle filtering (numerical methods);ARMA state process estimation;Gaussian distribution;Monte Carlo approach;Rao-Blackwellization approach;dynamic unknown parameter;extensive computer simulations;nonlinear functions;particle filtering method;signal processing;state-space model;static unknown parameter;Approximation methods;Autoregressive processes;Biological system modeling;Estimation;Mathematical model;Monte Carlo methods;State-space methods;ARMA processes;Rao-Blackwellization;particle filtering;state-space estimation},
  Owner                    = {iurteaga},
  Timestamp                = {2014-10-01}
}

There are many practical signal processing settings where a state-space model consists of a state described by an ARMA process that is observed via non-linear functions of the state. In this paper, we propose a particle filtering method for sequentially estimating the ARMA process in the presence of unknown parameters. In the considered problem, we have static and dynamic unknowns, and we show how to handle the static parameters so that the estimation of the state process does not degrade with time. We propose a new particle filter that approximates the posterior of all the unknowns by a Gaussian distribution, in combination with a Monte Carlo approach to the Rao-Blackwellization of the static parameters. We demonstrate the performance of the proposed method by extensive computer simulations.

@InProceedings{ip-Johnston2013,
  Title                    = {{Replication and optimization of hedge fund risk factor exposures}},
  Author                   = {Douglas E. Johnston and I{\~n}igo Urteaga and Petar M. Djuri\'{c}},
  Booktitle                = {2013 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  Year                     = {2013},
  Month                    = {May},
  Pages                    = {8712-8716},

  Abstract                 = {In this paper, we propose a novel approach for decomposing hedge fund returns onto observable risk factors. We utilize a vector stochastic-volatility model to extract the time-varying exposure of low frequency hedge fund returns on high frequency market data. We implement the estimation by using particle filtering and the concept of Rao-Blackwellization. With the latter, we remove all the static parameters of the model and thereby reduce the dimension of the parameter space for particle generation. Thus, we are able to obtain accurate estimates of the posterior distributions of the model states. For our model, this reduction is significant because the number of static parameters is large. We use the proposed model to analyze hedge fund performance and to optimally replicate hedge fund strategies economically. We demonstrate the validity and effectiveness of the method by computer simulations.},
  Doi                      = {http://dx.doi.org/10.1109/ICASSP.2013.6639367},
  ISSN                     = {1520-6149},
  Keywords                 = {investment;optimisation;particle filtering (numerical methods);risk management;stochastic processes;Rao-Blackwellization;computer simulations;hedge fund performance;hedge fund returns decomposition;hedge fund risk factor exposures optimization;hedge fund risk factor exposures replication;hedge fund strategies;high frequency market data;low frequency hedge fund returns;model states;observable risk factors;parameter space;particle filtering;particle generation;posterior distributions;static parameters;time-varying exposure;vector stochastic-volatility model;Bayes methods;Computational modeling;Mathematical model;Portfolios;Stochastic processes;Stock markets;Vectors;CAPM;VaR;beta;hedge fund;particle filtering;risk-management;stochastic volatility},
  Owner                    = {iurteaga},
  Timestamp                = {2014-02-10}
}

In this paper, we propose a novel approach for decomposing hedge fund returns onto observable risk factors. We utilize a vector stochastic-volatility model to extract the time-varying exposure of low frequency hedge fund returns on high frequency market data. We implement the estimation by using particle filtering and the concept of Rao-Blackwellization. With the latter, we remove all the static parameters of the model and thereby reduce the dimension of the parameter space for particle generation. Thus, we are able to obtain accurate estimates of the posterior distributions of the model states. For our model, this reduction is significant because the number of static parameters is large. We use the proposed model to analyze hedge fund performance and to optimally replicate hedge fund strategies economically. We demonstrate the validity and effectiveness of the method by computer simulations.

@InProceedings{ip-Urteaga2011,
  Title                    = {{AWARE: Activity AWARE network clustering for wireless sensor networks}},
  Author                   = {I{\~n}igo Urteaga and Na Yu and Nicholas Hubbell and Qi Han},
  Booktitle                = {IEEE Local Computer Networks},
  Year                     = {2011},
  Pages                    = {589-596},

  Bibsource                = {DBLP, http://dblp.uni-trier.de},
  Doi                      = {http://dx.doi.org/10.1109/LCN.2011.6115521},
  Owner                    = {iurteaga},
  Timestamp                = {2014-02-11}
}

@InProceedings{ip-Urteaga2010,
  Title                    = {{On the design of a scalable multimedia streaming system based on receiver-driven flow and congestion awareness}},
  Author                   = {I{\~n}igo Urteaga and Iraide Unanue and Javier Del Ser and Pedro J. S{\'a}nchez and Aitor Rodriguez},
  Booktitle                = {2010 International Conference on Signal Processing and Multimedia Applications (SIGMAP)},
  Year                     = {2010},
  Month                    = {July},
  Pages                    = {39-45},

  Abstract                 = {In this position paper we present the design of an end-to-end scalable content streaming system that optimizes the quality of experience of the end-user by allowing each client to retrieve a customized multimedia stream, based on both network and client states. By taking advantage of multimedia scalability, our proposed receiverdriven architecture performs a multilayered streaming, where each client is responsible for controlling the number of multimedia layers it demands from the server. Furthermore, the streaming system proposed herein implements both congestion and flow control mechanisms, which are also delegated to the receiver. In order to properly address both network and client states and restrictions, a set of specific metrics (Buffer State, Interarrival Jitter and Loss Event Rate) are utilized, which have been specifically designed to match the miscellaneous characteristics of heterogeneous networks and end devices. Built upon such metrics, we present a decision algorithm that jointly performs congestion and flow control, while maximizing inter-session fairness and end-user quality of experience. The proposed architecture combines different standard protocols while guaranteeing independence between components of the streaming system.},
  Day                      = {28},
  Keywords                 = {Measurement;Multimedia communication;Protocols;Real time systems;Servers;Static VAr compensators;Streaming media;Congestion control;Flow control;Scalable multimedia content;Streaming},
  Location                 = {Athens, Greece},
  Owner                    = {iurteaga},
  Timestamp                = {2014-02-11},
  Url                      = {http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=5742565&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D5742565}
}

In this position paper we present the design of an end-to-end scalable content streaming system that optimizes the quality of experience of the end-user by allowing each client to retrieve a customized multimedia stream, based on both network and client states. By taking advantage of multimedia scalability, our proposed receiverdriven architecture performs a multilayered streaming, where each client is responsible for controlling the number of multimedia layers it demands from the server. Furthermore, the streaming system proposed herein implements both congestion and flow control mechanisms, which are also delegated to the receiver. In order to properly address both network and client states and restrictions, a set of specific metrics (Buffer State, Interarrival Jitter and Loss Event Rate) are utilized, which have been specifically designed to match the miscellaneous characteristics of heterogeneous networks and end devices. Built upon such metrics, we present a decision algorithm that jointly performs congestion and flow control, while maximizing inter-session fairness and end-user quality of experience. The proposed architecture combines different standard protocols while guaranteeing independence between components of the streaming system.

@InProceedings{ip-Urteaga2009,
  Title                    = {{REDFLAG a Run-timE, Distributed, Flexible, Lightweight, And Generic fault detection service for data-driven wireless sensor applications}},
  Author                   = {I{\~n}igo Urteaga and Kevin Barnhart and Qi Han},
  Booktitle                = {IEEE International Conference on Pervasive Computing and Communications, 2009},
  Year                     = {2009},
  Month                    = {March},
  Pages                    = {1-9},

  Abstract                 = {Increased interest in Wireless Sensor Networks (WSNs) by scientists and engineers is forcing WSN research to focus on application requirements. Data is available as never before in many fields of study; practitioners are now burdened with the challenge of doing data-rich research rather than being data-starved. In-situ sensors can be prone to errors, links between nodes are often unreliable, and nodes may become unresponsive in harsh environments, leaving to researchers the onerous task of deciphering often anomalous data. Presented here is the REDFLAG fault detection service for WSN applications, a Run-timE, Distributed, Flexible, detector of faults, that is also Lightweight And Generic. REDFLAG addresses the two most worrisome issues in data-driven wireless sensor applications: abnormal data and missing data. REDFLAG exposes faults as they occur by using distributed algorithms in order to conserve energy. Simulation results show that REDFLAG is lightweight both in terms of footprint and required power resources while ensuring satisfactory detection and diagnosis accuracy. Because REDFLAG is unrestrictive, it is generically available to a myriad of applications and scenarios.},
  Doi                      = {http://dx.doi.org/10.1109/PERCOM.2009.4912766},
  Keywords                 = {distributed algorithms;fault diagnosis;wireless sensor networks;REDFLAG;abnormal data;data-driven wireless sensor applications;distributed algorithms;generic fault detection service;in-situ sensors;missing data;wireless sensor networks;Application software;Calibration;Data engineering;Distributed algorithms;Fault detection;Network topology;Runtime;Sensor phenomena and characterization;Signal processing;Wireless sensor networks},
  Owner                    = {iurteaga},
  Timestamp                = {2014-02-11}
}

Increased interest in Wireless Sensor Networks (WSNs) by scientists and engineers is forcing WSN research to focus on application requirements. Data is available as never before in many fields of study; practitioners are now burdened with the challenge of doing data-rich research rather than being data-starved. In-situ sensors can be prone to errors, links between nodes are often unreliable, and nodes may become unresponsive in harsh environments, leaving to researchers the onerous task of deciphering often anomalous data. Presented here is the REDFLAG fault detection service for WSN applications, a Run-timE, Distributed, Flexible, detector of faults, that is also Lightweight And Generic. REDFLAG addresses the two most worrisome issues in data-driven wireless sensor applications: abnormal data and missing data. REDFLAG exposes faults as they occur by using distributed algorithms in order to conserve energy. Simulation results show that REDFLAG is lightweight both in terms of footprint and required power resources while ensuring satisfactory detection and diagnosis accuracy. Because REDFLAG is unrestrictive, it is generically available to a myriad of applications and scenarios.